An AQMesh air quality monitoring sensor was installed at a monitoring site in East Hartford, Connecticut to evaluate its performance tracking gas, particulate and meteorology data over a year long time frame. Hourly data for O3, NO2, PM2.5, PM10, temperature and relative humidity were compared to reference monitors located at the same site. The full downloadable dataset used is located here Download CSV.
| Possible Configuration | Evaluated Configuration | Cost | Data Access | Power Supply | Considerations | Reference Monitors Compared |
|
Particulates: PM1, PM2.5, PM10
Gases: NO, NO2, O3, CO, CO2, TVOC, SO2, EtO Meteorology: WS/WD, Baro Pressure, Temp, RH |
Particulates: PM2.5, PM10 Gases: O3, NO2 Meteorology: Temp, RH |
Sensor: $6,663.00 Sonic Anemometer: $4,998.00 Solar Configuration: $1,967.00 Cellular Communications (Annual): $420.00 API or Dashboard: $1,260.00 |
Basic Download (CSV), API, Web Application (Data stored locally as backup only, inaccessible to customer) |
Smart Solar Pack, Rechargeable NiMH battery, Mains DC Power |
Time Resolution: 1 minute to 1 hour intervals Dimensions: 430 (H) x 220 (W) x 170 (L) mm (including antenna) Weight: 2-2.7 kg |
O3: Teledyne N400 NO2: Teledyne T500U PM2.5 & PM10: Teledyne API T640X Temp & RH: Climatronix & Vaisala via AutoMet 580 |
One AQMesh pod was installed at CT DEEP’s McAuliffe Park Ambient Air Monitoring Station following collocation requirements from Ambilabs: sited greater than 0.5m above roof level, 1 meter horizontally from reference inlets, and free of any obstructions that could impact the free movement of air. The AQMesh was powered by a solar panel and battery pack purchasable from Ambilabs, and wind speed and direction were provided through a Davis anemometer. Data from the AQMesh pod were streamed to a central DAS over a 3G cellular connection and averaged into 15-minute intervals. CT DEEP accessed data through an API feed. Data were averaged into hour-intervals for comparison with regulatory O3, NO2, PM2.5, PM10, and meteorological instrumentation. “Prescale” data fields were used in analysis. AQMesh offers linear regression analysis for collocated pods, but regressions were conducted by CT DEEP and applied to data manually to preserve corrected vs uncorrected data comparisons.
AQMesh hourly data for O3, NO2, PM2.5, PM10, temperature and relative humidity were compared by quarter to reference values (Q1: January-March, Q2: April-June, Q3: July-September, Q4: October-December).
*Outliers were defined from the AQMesh dataset using the IQR Method
(all data points more than 1.5 below the the lower bound quartile or
above the upper bound quartile).
PM10 readings were significantly impacted by the “fog affect” that typically occurs at dawn between the hours of 3-6am EST during summer months. AQMesh flags coarse particulate readings with the code “deliquescence” if measured RH values are above a threshold, however those flags were not largely applied in our dataset. During these fog events, PM10 values were reported at anomalously high values due to interference with the optical particle counter. These outliers were removed from the dataset using the IQR method, in which all data points 1.5 beyond the lower and upper bound quartiles were excluded. Due to the limitations of the OPCs, data from these fog windows- especially in summer months- would have to be excluded from a dataset. Several studies suggest a heatded inlet would decrease this humidity sensitivity, however that was not evaluated due to our solar configuration.
With outliers removed from the dataset, PM2.5 was the only parameter which fell within acceptable levels in comparison to EPA Air Sensor Performance Targets. PM2.5 shows significant correlation through linear regression models and acceptable offset and RMSE, with slope just outside of the target range. With outliers excluded, PM10 data show acceptable offset and RMSE, but with slope and R2 beyond target ranges. O3 data are just outside of targets for slope, R2 and RMSE, but display an acceptable offset. NO2 measurements showed weak to non-correlation with the regulatory instrument. Temperature and RH measurements in the pod were expected to vary slightly from station measurements due to the location of the sensors within the sensor housing. However, results show that temperature and RH correlate well with ambient measurements and can be expected to track near ambient levels. Data capture is 100% during the testing window, with no significant (greater than 7 minutes) power disruptions. The only data unavailable for analysis were PM2.5 and PM10 measurements flagged for deliquescence in the AQMesh DAS.